Matt Marx

Co-Founder @PhishFort

[Matt Marx @tehnlulz]

It's incredible how bad GPT 5.5 feels after using Fable. I loved 5.5 a week ago, but Fable was such a game changer and step up from everything else.

It's really apparent that closed groups that get access to these in the future models are going to be completely untouchable.

[Keir Starmer @Keir_Starmer]

In the early hours of this morning, I directed our Armed Forces to intercept a shadow fleet oil tanker attempting to pass through the English Channel.

This successful operation delivers yet another blow to Russia and reminds those fueling Putin's war in Ukraine that we will not let them hide.

I want to thank those involved, including our Armed Forces and law enforcement officers who keep this country safe 24 hours a day, 365 days a year.

[ClaudeDevs @ClaudeDevs]

As a result of a US government directive, we are suspending access to Claude Fable 5 for all users. You can continue to use all other Claude models.

Here’s what this means for you:

Across Claude products, new sessions will run on your selected default model or Opus 4.8, and existing Fable 5 sessions will end with an error.

On the Claude Platform, requests to Fable 5 will also return an error. Please update your integrations to other Claude models.

We know this is a disruption to your workflows; we appreciate your patience and support.

[Matt Marx @tehnlulz]

I want to believe

If this actually feels as good as it benchmarks, this will be special.

An inflection point

[Matt Marx @tehnlulz]

This looks very, very cool 😎

[cat @_catwu]

Excited to share our most powerful new Claude Code feature: dynamic workflows!

Mention "workflow" in a prompt and Claude will dynamically create an orchestration plan that it strictly follows, allowing you to confidently trust that every stage happens in the right order even across 100s of agents.

[Matt Marx @tehnlulz]

👀

[Riley Brown @rileybrown]

It will be a big day from both teams...

[Matt Marx @tehnlulz]

Had a chance to play around with @getopenwork this weekend.

I don't know if most people are aware just how good local models are now but for a while I've been looking for a way to do Claude Cowork / Codex work stuff (not for SWE tasks) locally with open source stack.

Openwork is really nice for this! I connect it to @lmstudio running locally and run the qwen3.6-35b-a3b-mtp quants from @UnslothAI and it works well on my MBP (@ around 75 tok/sec).

Openwork is great, it supports browser use straight out of the box. It does a good job as a harness for work stuff. It's early and has some stuff missing/bugs I'm sure the team will work on as they grow but it's possible to do what used to be possible only with closed source, privacy unfriendly frontier labs.

For so long you had to choose between using a local private model which kinda sucked and using frontier lab model & software ecosystem. It's awesome to see open source products and solutions which are actually usable. Nice work @getopenwork team!

[Matt Marx @tehnlulz]

@ronaldlangeveld 100% and it shows - just need to spend a couple days in SG and everything works, is clean, low crime etc

[Matt Marx @tehnlulz]

Wow.

Singapore can be so forward-thinking sometimes. I wonder if smaller, faster moving countries like Singapore are going to see outsized returns with AI (like smaller companies are compared with the slower, bigger, bureaucratic incumbents).

[Gabriel Chua @gabrielchua]

OpenAI for Singapore 🇸🇬

[GitHub @github]

We are investigating unauthorized access to GitHub’s internal repositories. While we currently have no evidence of impact to customer information stored outside of GitHub’s internal repositories (such as our customers’ enterprises, organizations, and repositories), we are closely monitoring our infrastructure for follow-on activity.

[Matt Marx @tehnlulz]

😂

[ChatGPT @ChatGPTapp]

Prompt:

“Redraw the attached image in the most clumsy, scribbly, and utterly pathetic way possible. Use a white background, and make it look like it was drawn in MS Paint with a mouse. It should be vaguely similar but also not really, kind of matching but also off in a confusing, awkward way, with that low-quality pixel-by-pixel feel that really emphasizes how ridiculously bad it is. Actually, you know what, whatever, just draw it however you want.”

[Matt Marx @tehnlulz]

Skeptical but would be nice

[Theo - t3.gg @theo]

I legitimately believe xAI might have a crazy comeback

[Matt Marx @tehnlulz]

One way you can tell how far models like GPT-5.5 have come is to reflect on how difficult it is to tell the difference between 5.5 and 5.4. I have to rely on obscure benchmarks and hours of working with the model to ‘get’ it.

[Matt Marx @tehnlulz]

@threepointone I'm curious how it's changed your role at CF @threepointone? Do you have any talks where you discuss this? The best execs are all getting hands on now, that's clear. But interested in how you see things shaping out

[Matt Marx @tehnlulz]

@threepointone Sammeeee. I think one of the things I love the most about this era is that stuff like writing a custom GPU kernel or building some ungodly performant stuff in Rust - stuff I just never had enough time to learn and have no XP in that I can now do

Grateful to be alive rn

[Matt Marx @tehnlulz]

Kinda sad to see sentiment turn sour toward @AnthropicAI.

There’s a lesson here about the importance of transparency and communication with your users

[Matt Marx @tehnlulz]

😬😬😬

[Chris @ChrissGPT]

Tomorrow will be the biggest day for OpenAI since GPT 3.5

[Matt Marx @tehnlulz]

@SteveMoraco It was a glorious time

[Matt Marx @tehnlulz]

Another incredible local model release. If the frontier labs ever decide to stop blessing us with incredible but financially unsustainable subscriptions, I will still be able to limp along with coding work locally thanks to Qwen.

I've loved and used (and fine tuned) this model family since their first release, incredible to see how far they have come. Thank you @Alibaba_Qwen!

[Qwen @Alibaba_Qwen]

🚀 Meet Qwen3.6-27B, our latest dense, open-source model, packing flagship-level coding power!

Yes, 27B, and Qwen3.6-27B punches way above its weight. 👇

What's new:
🧠 Outstanding agentic coding — surpasses Qwen3.5-397B-A17B across all major coding benchmarks
💡 Strong reasoning across text & multimodal tasks
🔄 Supports thinking & non-thinking modes
✅ Apache 2.0 — fully open, fully yours

Smaller model. Bigger results. Community's favorite. ❤️
We can't wait to see what you build with Qwen3.6-27B! 👀

🔗👇
Blog: https://t.co/P2Zx7FwMxB
Qwen Studio: https://t.co/c4vm4LuZrU
Github: https://t.co/zKDEbv0R4U
Hugging Face:
https://t.co/N67hyzxvfr
https://t.co/SSdtbWRDap
ModelScope:
https://t.co/xODf1pj9kw
https://t.co/xXhoqlJ2AB

[Matt Marx @tehnlulz]

@chrisgpt Got this just before it disappeared:
https://t.co/pIOMcd5ZbQ

[Matt Marx @tehnlulz]

aaaand it's gone. I did manage to fire off one prompt. Looks great at design now!

Design and code a high-end, ultra-modern, single-file HTML/Tailwind CSS landing page for 'https://t.co/8jiA7vED59,' a cutting-edge cybersecurity brand specializing in anti-phishing and brand protection. The aesthetic must be 'Cyber-Noir meets Apple Pro'—use a deep midnight-black background (#050505) with subtle, animated mesh gradients in electric indigo and deep violet flowing in the background. Incorporate a heavy use of glassmorphism (frosted glass effects), ultra-thin borders with subtle glows, and high-contrast typography using a sleek sans-serif like Inter or Geist. The hero section must feature a massive, bold headline with a dramatic text gradient, a subheadline with perfect tracking, and a glowing 'Get Protected' CTA button that has a pulsing hover effect. Include a 'Live Threat Map' placeholder section using a dark, stylized grid pattern, a feature grid using bento-box style layouts with micro-interactions, and a high-tech 'Trust Bar' with monochromatic logos. Use Lucide icons for a crisp look, ensure every element has smooth CSS transitions, add a subtle grain texture overlay to the entire body for a premium cinematic feel, and make sure the layout is fully responsive, mobile-first, and feels like a $100k custom-coded boutique agency site.